update links to /book instead of /ed4

README.md

@@ -1,6 +1,6 @@
 # Computational Cognitive Neuroscience Simulations
 
-This repository contains the neural network simulation models for the [CCN Textbook](https://CompCogNeuro.org), managed on a [GitHub Repository](https://github.com/CompCogNeuro/ed4).
+This repository contains the neural network simulation models for the [CCN Textbook](https://CompCogNeuro.org), managed on a [GitHub Repository](https://github.com/CompCogNeuro/book).
 
 To run these simulations on your computer, it is easiest to download the full set of executable programs for the type of computer you are using (Apple Mac, Microsoft Windows, or Linux):
 

ch2/neuron/README.md

@@ -29,7 +29,7 @@ Here is a quick overview of each of the variables -- we'll go through them indiv
 
 # Spiking Behavior
 
-The default parameters that you just ran show the spiking behavior of a neuron. This is implementing a modified version of the Adaptive Exponential function (see [CCN Textbook](https://github.com/CompCogNeuro/ed4)) or AdEx model, which has been shown to provide a very good reproduction of the firing behavior of real cortical pyramidal neurons. As such, this is a good representation of what real neurons do. We have turned off the exponential aspect of the AdEx model here to make parameter manipulations more reliable -- a spike is triggered when the membrane potential Vm crosses a simple threshold of .5. (In contrast, when exponential is activated (you can find it in the `SpikeParams`), the triggering of a spike is more of a dynamic exponential process around this .5 threshold level, reflecting the strong nonlinearity of the sodium channels that drive spiking.)
+The default parameters that you just ran show the spiking behavior of a neuron. This is implementing a modified version of the Adaptive Exponential function (see [CCN Textbook](https://github.com/CompCogNeuro/book)) or AdEx model, which has been shown to provide a very good reproduction of the firing behavior of real cortical pyramidal neurons. As such, this is a good representation of what real neurons do. We have turned off the exponential aspect of the AdEx model here to make parameter manipulations more reliable -- a spike is triggered when the membrane potential Vm crosses a simple threshold of .5. (In contrast, when exponential is activated (you can find it in the `SpikeParams`), the triggering of a spike is more of a dynamic exponential process around this .5 threshold level, reflecting the strong nonlinearity of the sodium channels that drive spiking.)
 
 At the broadest level, you can see the periodic purple spikes that fire as the membrane potential gets over the firing threshold, and it is then reset back to the rest level, from which it then climbs back up again, to repeat the process again and again. Looking at the overall rate of spiking as indexed by the spacing between spikes (i.e., the *ISI* or inter-spike-interval), you can see that the spacing increases over time, and thus the rate decreases over time.  This is due to the **adaptation** property of the AdEx model -- the spike rate adapts over time.
 

ch3/inhib/README.md

@@ -51,7 +51,7 @@ A more intuitive (but somewhat inaccurate in the details) way of understanding t
 
 # Roles of Feedforward and Feedback Inhibition
 
-Next we assess the importance and properties of the feedforward versus feedback inhibitory projections by manipulating their relative strengths. The control panel has two parameters that determine the relative contribution of the feedforward and feedback inhibitory pathways: `FFinhibWtScale` applies to the feedforward weights from the input to the inhibitory units, and `FBinhibWtScale` applies to the feedback weights from the hidden layer to the inhibitory units. These parameters (specifically the .rel components of them) uniformly scale the strengths of an entire projection of connections from one layer to another, and are the arbitrary `WtScale.Rel` (r_k) relative scaling parameters described in *Net Input Detail* Appendix in [CCN TExtbook](https://github.com/CompCogNeuro/ed4).
+Next we assess the importance and properties of the feedforward versus feedback inhibitory projections by manipulating their relative strengths. The control panel has two parameters that determine the relative contribution of the feedforward and feedback inhibitory pathways: `FFinhibWtScale` applies to the feedforward weights from the input to the inhibitory units, and `FBinhibWtScale` applies to the feedback weights from the hidden layer to the inhibitory units. These parameters (specifically the .rel components of them) uniformly scale the strengths of an entire projection of connections from one layer to another, and are the arbitrary `WtScale.Rel` (r_k) relative scaling parameters described in *Net Input Detail* Appendix in [CCN Textbook](https://github.com/CompCogNeuro/book).
 
 * Set `FFInhibWtScale` to 0, effectively eliminating the feedforward excitatory inputs to the inhibitory neurons from the input layer (i.e., eliminating feedforward inhibition). 
 
@@ -65,7 +65,7 @@ These exercises should help you to see that a combination of both feedforward an
 
 ## Time Constants and Feedforward Anticipation
 
-We just saw that feedforward inhibition is important for anticipating and offsetting the excitation coming from the inputs to the hidden layer. In addition to this feedforward inhibitory connectivity, the anticipatory effect depends on a difference between excitatory and inhibitory neurons in their rate of updating, which is controlled by the `Dt.GTau` parameters `HiddenGTau` and `InhibGTau` in the control panel (see [CCN Textbook](https://github.com/CompCogNeuro/ed4), Chapter 2). As you can see, the excitatory neurons are updated at tau of 40 (slower), while the inhibitory are at 20 (faster) -- these numbers correspond roughly to how many cycles it takes for a substantial amount of change happen. The faster updating of the inhibitory neurons allows them to more quickly become activated by the feedforward input, and send anticipatory inhibition to the excitatory hidden units before they actually get activated.
+We just saw that feedforward inhibition is important for anticipating and offsetting the excitation coming from the inputs to the hidden layer. In addition to this feedforward inhibitory connectivity, the anticipatory effect depends on a difference between excitatory and inhibitory neurons in their rate of updating, which is controlled by the `Dt.GTau` parameters `HiddenGTau` and `InhibGTau` in the control panel (see [CCN Textbook](https://github.com/CompCogNeuro/book), Chapter 2). As you can see, the excitatory neurons are updated at tau of 40 (slower), while the inhibitory are at 20 (faster) -- these numbers correspond roughly to how many cycles it takes for a substantial amount of change happen. The faster updating of the inhibitory neurons allows them to more quickly become activated by the feedforward input, and send anticipatory inhibition to the excitatory hidden units before they actually get activated.
 
 * To verify this, click on Defaults, set `InhibGTau` to 40 (instead of the 20 default), and then Run. 
 
@@ -119,7 +119,7 @@ This reduces the amount of inhibition on the excitatory neurons. Note that this
 
 # Exploration of FFFB Inhibition
 
-You should run this section after having read the *FFFB Inhibition Function* section of the [CCN Textbook](https://github.com/CompCogNeuro/ed4).
+You should run this section after having read the *FFFB Inhibition Function* section of the [CCN Textbook](https://github.com/CompCogNeuro/book).
 
 * Reset the parameters to their default values using the `Defaults` button, click the `BidirNet` on to use that, and then Test to get the initial state of the network. This should reproduce the standard activation graph for the case with actual inhibitory neurons.
 

ch4/self_org/README.md

@@ -2,15 +2,15 @@ Back to [All Sims](https://github.com/CompCogNeuro/sims) (also for general info
 
 # Introduction
 
-This model illustrates how self-organizing learning emerges from the interactions between the following factors (as discussed in the *Learning* Chapter of the  [CCN Textbook](https://github.com/CompCogNeuro/ed4) ):
+This model illustrates how self-organizing learning emerges from the interactions between the following factors (as discussed in the *Learning* Chapter of the  [CCN Textbook](https://github.com/CompCogNeuro/book) ):
 
 * **Inhibitory competition** -- only the most strongly driven neurons get over the inhibitory threshold, and can get active. These are the ones whose current synaptic weights best fit ("detect") the current input pattern.
 
 * **Rich get richer** positive feedback loop -- due to the nature of the learning function, only those neurons that actually get active are capable of learning (when receiver activity y = 0, then xy = 0 too, and the XCAL dWt function is 0 at 0). Thus, the neurons that already detect the current input the best are the ones that get to further strengthen their ability to detect these inputs. This is the essential insight that Hebb had with why the Hebbian learning function should strengthen an "engram".
 
 * **homeostasis** to balance the positive feedback loop -- if left unchecked, the rich-get-richer dynamic ends up with a few units dominating everything, and as a result, all the inputs get categorized into one useless, overly broad category ("everything"). The homeostatic mechanism in BCM helps fight against this by raising the floating threshold for highly active neurons, causing their weights to decrease, and restoring a balance. Similarly, under-active neurons experience net weight increases that get them participating and competing more effectively.
 
-The net result is the development of a set of neural detectors that relatively evenly cover the space of different inputs patterns, with systematic categories that encompass the statistical regularities. For example, cats like milk, and dogs like bones, and we can learn this just by observing the reliable co-occurrence of cats with milk and dogs with bones. This kind of reliable co-ocurrence is what we mean by "statistical regularity". See *Hebbian Learning* Appendix in the [CCN Textbook](https://github.com/CompCogNeuro/ed4) for a very simple illustration of why Hebbian-style learning mechanisms capture patterns of co-occurrence. It is really just a variant on the basic maxim that "things that fire together, wire together".
+The net result is the development of a set of neural detectors that relatively evenly cover the space of different inputs patterns, with systematic categories that encompass the statistical regularities. For example, cats like milk, and dogs like bones, and we can learn this just by observing the reliable co-occurrence of cats with milk and dogs with bones. This kind of reliable co-ocurrence is what we mean by "statistical regularity". See *Hebbian Learning* Appendix in the [CCN Textbook](https://github.com/CompCogNeuro/book) for a very simple illustration of why Hebbian-style learning mechanisms capture patterns of co-occurrence. It is really just a variant on the basic maxim that "things that fire together, wire together".
 
 In this exploration, the network learns about a simple world that consists purely of horizontal and vertical lines, with these lines appearing always in combination with other lines. The clear objective of self-organizing learning in this case is to extract the underlying statistical regularity that these lines exist as reliable collections of pixels, and it would be much more efficient to encode this world in terms of the lines, instead of individual pixels.
 

doc.go

@@ -3,7 +3,7 @@
 // license that can be found in the LICENSE file.
 
 /*
-	Package sims are the neural network simulation models for the [CCN Textbook](https://github.com/CompCogNeuro/ed4).
+	Package sims are the neural network simulation models for the [CCN Textbook](https://github.com/CompCogNeuro/book).
 
 These models are implemented in the new *Go* (golang) version of
 [emergent](https://github.com/emer/emergent), with Python versions